Does relative valuation work for banks?

We study the distribution and properties of valuation errors yielded by banking industry multiples for European and U.S. banks. The results highlight that stock-market multiples are best suited for U.S. institutions, and that a two-year-forward P/E is the most precise metric. Contrary to practitioner beliefs, P/tangible book value is less meaningful than P/BV. Multiples are less accurate for small commercial banks than for large ones, and for investment banks than for retail banks. We investigate whether large positive errors lead to one-year positive price performances and negative errors to negative price changes, and find that the forward P/E loses its predictive ability in comparison with historical multiples. Testing three investment strategies, we find that bank multiples can be profitably used in portfolio choices.